os: add os.devNull

[lpinca]

Provides the platform-specific file path of the null device.

[lpinca]

I'm not sure if it makes sense but it is similar to path.sep or path.delimiter. In Python this is available via os.devnull.

[RaisinTen]

https://docs.python.org/3/library/os.html#os.devnull says:

'/dev/null' for POSIX, 'nul' for Windows.

So, why isn't this just 'nul' for Windows?

[lpinca]

Because if you do something like

fs.writeFileSync(path.devnull, 'foo');

it does not work as expected.

[lpinca]

See also nodejs/node-v0.x-archive#9271

[RaisinTen]

Should we add a test that attempts to write something to the file and then asserts that the file size is 0?

[mscdex]

I'm -0 on adding this, but if anything it should probably live on os and not path since path is about manipulating the paths themselves and retrieving a platform-specific "devnull" path/device is more like a os.tmpdir() thing.

[jasnell]

Dropping this off os definitely makes the most sense. I do have questions about the utility of this, however... specifically, what would be the purpose of writing explicitly to nul as opposed to simply just not writing anything?

[lpinca]

@mscdex I thought about that but the same argument can be done for path.sep and path.delimiter. At least there is some consistency. Also the Pyhton implementation lives in the {nt,posix}path module and is then reexported in the os module. It is also available as os.path.devnull.

I don't like to have it on both os and path but we could that if there is consensus.

[lpinca]

@jasnell tests, I used it here https://github.com/lpinca/fritzbox-checksum/blob/fe8311da86062a3c748b9bc3e69abdb28bb9003b/test/test.js#L83.

It could definitely be done differently but why if a null device exists?

[jasnell]

Hmmm.. ok. Thinking about it further, hanging it off path would be ok for me so long as it's just a constant similar to separator. If there's even a theoretical need to have to query the os to find the null path, then it should live in os.

[devsnek]

+1 to a method in os, even if the implementations just return a static string.

[lpinca]

@devsnek I'm fine with moving it to os but why a method? There is also os.EOL.

[mscdex]

I thought about that but the same argument can be done for path.sep and path.delimiter

Those are components of a path, so to me it makes sense for them to live on path.

[lpinca]

Fair enough, moved it to os.devnull().

[sindresorhus]

I think it should be os.devNull (note the camel casing). It's two words: "dev null". It should also be a property, like os.EOL.

[jasnell]

I think it should be os.devNull (note the camel casing). It's two words: "dev null". It should also be a property, like os.EOL.

I agree. I'm ok with the method and the current naming and definitely wouldn't block it (which is why I signed off) but my preference would be a simple property.

[lpinca]

There is os.loadavg(), os.homedir(), os.freemem(), os.totalmem() which all could use camel casing.

os.tmpDir() was explicitly renamed to os.tmpdir() in 3fe6aba. The alias was deprecated in 5e5ec2c and then removed in ab4115f.

os.endianness() also returns just a string constant but it's a method and not a property.

I'm just trying to make things consistent.

[jasnell]

Still LGTM!

[cjihrig]

LGTM

[RaisinTen]

LGTM 👍

[bl-ue]

Any reason not to store os.devNull in a separate variable? I'm not getting it.

Suggested change

	const devNull = os.devNull;
	if (common.isWindows) {
	assert.strictEqual(devNull, '\\\\.\\nul');
	} else {
	assert.strictEqual(devNull, '/dev/null');
	}
	if (common.isWindows) {
	assert.strictEqual(os.devNull, '\\\\.\\nul');
	} else {
	assert.strictEqual(os.devNull, '/dev/null');
	}

[lpinca]

Consistency with

node/test/parallel/test-os.js

Lines 186 to 191 in 1fc60ab

	const EOL = os.EOL;
	if (common.isWindows) {
	assert.strictEqual(EOL, '\r\n');
	} else {
	assert.strictEqual(EOL, '\n');
	}

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/38375/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/38404/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/38430/

[lpinca]

Landed in 152d675